Public lecture by Prof. Tillmann Vierkant (University of Edinburgh) on AI and its risks to epistemic and moral agency.
In the literature authors traditionally diagnose a responsibility gap for responsible AI use. They argue that because of the opaque nature of AI decision making and the absence of direct control over the behaviour of autonomous systems it is often impossible to establish who is responsible when things go wrong and that this is intuitively problematic. Here I argue that this focus on epistemic opacity and absence of direct control is a red herring. This is because, as recent cognitive science developments show, we face surprisingly similar challenges when it comes to establishing whether humans have the right knowledge and the right control to be responsible for their actions. Philosophers who are impressed by this challenge to human responsibility have recently developed new instrumentalist views, which emphasize the forward-looking and flexible role of agency cultivation for responsibility. These views are not in the same way dependent on knowledge and control and therefore provide a plausible route to think about responsible AI use. But while instrumentalist views show great promise there is one underdiscussed problem in the case of AI that such accounts also struggle with. Responsibility practices for humans are a high stakes enterprise that we cannot escape if we want to be accepted as fully responsible agents. I argue that neither the AI nor any human component of the system faces similar existential vulnerabilities. I conclude that this invulnerability to moral challenge is the most significant responsibility gap when it comes to AI.
This event is free and open to the general public interested in the ethics and philosophy of AI.